Enhancement of the Deformable Image Registration Accuracy Using Image Modification of MV CBCT

Kim, Min Joo; Chang, Jina; Park, So Hyun; Kim, Tae Ho; Kang, Young Nam; Suh, Tae Suk

Abstract

To perform the Adaptive Radiation Therapy (ART), a high degree of deformable registration accuracy is essential. The purpose of this study is to identify whether the change of MV CBCT intensity can improve registration accuracy using predefined modification level and filtering process. To obtain modification level, the cheese phantom images was acquired from both kilovoltage CT (kappaV CT), megavoltage cone-beam CT (MV CBCT). From the cheese phantom images, the modification level of MV CBCT was defined from the relationship between Hounsfield Units (HUs) of kappaV CT and MV CBCT images. 'Gaussian smoothing filter' was added to reduce the noise of the MV CBCT images. The intensity of MV CBCT image was changed to the intensity of the kappaV CT image to make the two images have the same intensity range as if they were obtained from the same modality. The demon deformable registration which was efficient and easy to perform the deformable registration was applied. The deformable lung phantom which was intentionally created in the laboratory to imitate the changes of the breathing period was acquired from kappaV CT and MV CBCT. And then the deformable lung phantom images were applied to the proposed method. As a result of deformable image registration, the similarity of the correlation coefficient was used for a quantitative evaluation of the result was increased by 6.07% in the cheese phantom, and 18% in the deformable lung phantom. For the additional evaluation of the registration of the deformable lung phantom, the centric coordinates of the mark which was inserted into the inner part of the phantom were measured to calculate the vector difference. The vector differences from the result were 2.23, 1.39 mm with/without modification of intensity of MV CBCT images, respectively. In summary, our method has quantitatively improved the accuracy of deformable registration and could be a useful solution to improve the image registration accuracy. A further study was also suggested in this paper.